Assessing the accuracy of a spatial model of habitat suitability for Calypso bulbosa

Sammanfattning: Calypso bulbosa is a rare and visually striking orchid that grows in older mesic to moist forests in the northern half of Sweden. C. bulbosa is red listed as a threatened species (Vulnerable, VU) with a reduction in numbers linked to modern forestry practices and exacerbated by the warming climate. The species is protected in Sweden and appears in two appendices of the EU’s Species and Habitats Directive (EEA, 2016). To properly monitor the population and its response to a warming climate, an accurate model of species distribution is required. This data will be crucial in sustainable forestry management and conservation strategy development. In 2022 a model was developed by SLU Artdatabanken (English: Swedish Species Information Centre) at the Swedish University of Agricultural Sciences. This model suggested a number of potential locations for C. bulbosa presence in the alpine region, suggesting that current estimates of a C. bulbosa population of around 1000 individuals in the region may not be entirely accurate. This thesis tested the 2022 model’s predictive capability by visiting a number of areas in the Swedish alpine region with varying probability levels for C. bulbosa presence, with a special focus on “hotspots” that the model deemed to have a high likelihood of C. bulbosa occurrence. The result of these visits was then tested statistically against expected occurrence values from the 2022 model. Over the course of the fieldwork, two new C. bulbosa populations were discovered in the alpine region, one of which being the largest in the region discovered in Sweden in the 21st century thus far. In total, 1190 individuals of C. bulbosa were observed in the alpine region over the course of the fieldwork, a figure which surpasses the currently estimated population for the entire region. Despite these successes, the model itself did not appear to have a good predictive power in the alpine region at the hectare scale, with an AUC=0.445 making the model’s predictive capacity no better than random chance in this regard. The presence/absence data collected from the period of fieldwork can now be used to potentially improve the model itself, thereby improving the accuracy of C. bulbosa population estimations in Sweden. Accurate modelling and population tracking are crucial tools for responsible ecological management and policymaking.